The K-means algorithm is very fast and due to its speed and ease of implementation has become one of the most popular clustering techniques. The algorithm is sensitive to the initial choice of cluster centers and there is an extensive research on initialization methods (Lozano et al., 1999 and Erisoglu et al., 2011). In particular, it can be noted that the K-means algorithm experiences difficulties when the initial center “seeds” are chosen so that there are multiple “seeds” in the same cluster and some clusters are overlooked. A very common and the most simple way of dealing with this complication is to restart the algorithm several times picking the center “seeds” at random.